Risk factors for carbapenem-resistant Klebsiella pneumoniae infection relative to two types of control patients: a systematic review and meta-analysis

Background Studies on risk factors for carbapenem-resistant Klebsiella pneumoniae (CRKP) infection have provided inconsistent results, partly due to the choice of the control group. We conducted a systematic review and meta-analysis to assess the risk factors for CRKP infection by comparing CRKP-infected patients with two types of controls: patients infected with carbapenem-susceptible Klebsiella pneumoniae (comparison 1) or patients not infected with CRKP (comparison 2). Methods Data on potentially relevant risk factors for CRKP infection were extracted from studies indexed in PubMed, EMBASE, Web of Science or EBSCO databases from January 1996 to April 2019, and meta-analyzed based on the outcomes for each type of comparison. Results The meta-analysis included 18 studies for comparison 1 and 14 studies for comparison 2. The following eight risk factors were common to both comparisons: admission to intensive care unit (ICU; odds ratio, ORcomparison 1 = 3.20, ORcomparison 2 = 4.44), central venous catheter use (2.62, 3.85), mechanical ventilation (2.70, 4.78), tracheostomy (2.11, 8.48), urinary catheter use (1.99, 0.27), prior use of antibiotic (6.07, 1.61), exposure to carbapenems (4.16, 3.84) and exposure to aminoglycosides (1.85, 1.80). Another 10 risk factors were unique to comparison 1: longer length of hospital stay (OR = 15.28); prior hospitalization (within the previous 6 months) (OR = 1.91); renal dysfunction (OR = 2.17); neurological disorders (OR = 1.52); nasogastric tube use (OR = 2.62); dialysis (OR = 3.56); and exposure to quinolones (OR = 2.11), fluoroquinolones (OR = 2.03), glycopeptides (OR = 3.70) and vancomycin (OR = 2.82). Conclusions Eighteen factors may increase the risk of carbapenem resistance in K. pneumoniae infection; eight factors may be associated with both K. pneumoniae infections in general and CRKP in particular. The eight shared factors are likely to be ‘true’ risk factors for CRKP infection. Evaluation of risk factors in different situations may be helpful for empirical treatment and prevention of CRKP infections.


Background
Carbapenem-resistant Gram-negative bacteria, mainly Klebsiella pneumoniae, are an emerging cause of healthcare-associated infections that pose a significant threat to public health [1]. The percentage of K. pneumoniae infections resistant to carbapenems continues to rise [2,3], with proportions exceeding 50% in parts of the Eastern Mediterranean and Europe [1,2]. K. pneumoniae carbapenemase originated in the northeastern USA in the early 2000s, but rapidly disseminated to other regions worldwide [4].
Carbapenem-resistant K. pneumoniae (CRKP) infection is difficult to treat since carbapenems are often considered last-resort antibiotics for severe K. pneumoniae infections. The most important genes that can confer carbapenem resistance (via carbapenemases) are present in K. pneumoniae, rendering almost all available treatment options ineffective [2]. Mortality rates reach 33-50% among CRKP-infected patients in different regions of the world [5], significantly higher than mortality caused by infection with carbapenem-susceptible K. pneumoniae (CSKP) [1]. Preventing CRKP infection is therefore important not only to avoid poor prognosis and even death, but also to prevent widespread transmission of carbapenem resistance through mobile genetic elements [6,7].
Numerous studies have assessed risk factors for CRKP infection with different and sometimes even contradictory conclusions. A previous meta-analysis attempted to address this inconsistency [8] but did not take into consideration that different studies often use different control (reference) groups. The appropriate selection of the control group in the analysis of risk factors for antibioticresistant pathogen infections depends on the specific research question [9][10][11][12]. In studies analyzing risk factors for CRKP infection, two control groups are most often selected: patients infected with CSKP or patients without CRKP infection. The comparison of CRKP-infected with CSKP-infected patients may allow the identification of risk factors for carbapenem-resistant infections, although the results may be overestimated. In contrast, the comparison of CRKP-infected individuals with patients without CRKP infection may help to identify risk factors associated with both K. pneumoniae infections in general and CRKP in particular. Risk factors that are significant in both comparisons can be considered 'true' risk factors for CRKP infection [11,12].
Thus, we performed a systematic review and metaanalysis to clarify risk factors for CRKP infection relative to infection with CSKP (comparison 1) or to the absence of CRKP infection (comparison 2). This design, similar to a case-control-control study, aimed to compare the results of the two analyses and their different implications for the clinical practice, allowing the identification of the likely true risk factors for CRKP infection.

Methods
This meta-analysis was conducted according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [13].

Search strategy
Two authors (H.Y.Z. and Z.Y.) searched for relevant studies in PubMed, EMBASE, Web of Science and EBSCO databases that were published from January 1996 to April 2019. The search terms included "Klebsiella pneumoniae" AND ("carbapenem-resistant" OR "imipenem-resistant" OR "meropenem-resistant" OR "ertapenem-resistant" OR "carbapenemase-producing" OR "Klebsiella pneumoniae carbapenemase") AND ("risk factors" OR "risk" OR "factors"). Only studies published in English were considered. Reference lists in selected articles and relevant review articles were manually searched to identify additional studies.

Inclusion and exclusion criteria
Studies were included if they met the following criteria: (1) case-control or cohort study design, whether prospective or retrospective; (2) the risk factors for CRKP infection were reported; (3) either comparison 1 or comparison 2 was made; (4) CRKP and CSKP were classified based on K. pneumoniae isolate identification and tests for resistance to carbapenem (imipenem, meropenem, or ertapenem) involving well-defined microbiological methods; and (5) infection was explicitly defined. The inclusion criterion (3) led us to exclude studies comparing patients infected with carbapenemase-producing K. pneumoniae (CPKP) with controls without such infection, since such controls may have been infected with carbapenem-resistant, noncarbapenemase-producing K. pneumoniae. Studies were also excluded if they had the format of a report, review, comment, meeting abstract or letter to the editor; or if they reported insufficient data to assess outcomes.

Data extraction
Two authors (H.Y.Z. and W.M.Z.) independently evaluated and extracted data from the included studies using a predefined, standardized protocol. The extracted data on general characteristics of studies included the first author's name, year of publication, journal of publication, country, study period, study design and setting, type of inter-group comparison, sample size, average age, and sex distribution. Potential risk factors were included in the meta-analysis only if at least three studies examined them and those studies reported the numbers of individuals in each comparison group. Disagreements about extracted data were resolved through discussion.

Quality assessment
Two authors (W.M.Z. and Z.Y.) independently evaluated the quality of each study using the Newcastle-Ottawa Scale (NOS), a scale for assessing the quality of published non-randomized studies in meta-analyses [14]. The scale contains eight items, categorized into three dimensions: selection, comparability, and outcome (cohort studies) or exposure (case-control studies) [14]. We developed a NOS-based scale ranging from 0 to 9 points: studies scoring 0-4 points were defined as low quality, while those scoring 5-9 points were defined as high quality. Differences were resolved by consensus.

Statistical analysis
The meta-analysis was performed using RevMan 5.2 software provided by The Cochrane Collaboration (Copenhagen: The Nordic Cochrane Centre, 2014). Pooled odds ratios (ORs) and 95% confidence intervals (CIs) were calculated for all outcomes. The Z-test was used to determine the significance of the pooled OR, and the results were considered statistically significant when P < 0.05. Statistical heterogeneity among studies was assessed using a chi-squared test in which P < 0.10 was taken as the threshold for significant heterogeneity, or by calculating I 2 value, with I 2 > 50% considered evidence of heterogeneity [15]. Depending on the assessed heterogeneity, the Mantel-Haenszel fixed-or randomeffect methods were used to meta-analyze the outcomes.
Publication bias was quantitatively analyzed using Egger's test in STATA software version 12.0 (College Station, TX: StataCorp LP) [16], and the results were considered statistically significant when P < 0.05. Sensitivity analyses were conducted by omitting studies one by one, and the P values of pooled ORs were compared. The results were considered robust when the P values were not substantially different.

Study characteristics
The main characteristics of the 18 studies included in comparison 1 are presented in Table 1 (3), case-casecontrol (1), nested case-control (1), and prospective cohort (1). The comparison and reference groups were matched in 11 studies. All but three studies enrolled patients from a single center, and six studies enrolled only patients in the intensive care unit (ICU).
The main characteristics of the 14 studies included in comparison 2 are presented in Table 2. These studies were published from 2012 to 2019, and involved 893 patients with CRKP infection and 3073 without CRKP infection from six countries: Italy (6), USA (2), Greece (2), Turkey (2), Israel (1), and China (1). The designs of the studies were case-control (6), retrospective cohort (4), prospective cohort (2), case-case-control (1), and case-cohort (1). In six of these studies the comparison and reference groups were matched. All but one study enrolled patients from a single center and three studies involved only patients in the ICU.

Quality assessment
All studies in the review were judged to be of high quality based on NOS assessment. The 18 studies in comparison 1 scored an average of 7 (range 5-8) ( Table 1). The 14 studies in comparison 2 scored an average of 6 (range 5-8) ( Table 2). Table 3 shows the risk factors for CRKP infection for this comparison, as well as the heterogeneity in the meta-analysis. All 43 risk factors were dichotomous variables except for the following continuous variables: length of hospital stay (LOS), length of ICU stay, and Acute Physiology and Chronic Health Evaluation (APA-CHE) II score on ICU admission. Of the 43 factors, the following 18 were statistically significant: longer LOS, prior hospitalization (within the previous 6 months), admission to ICU, renal dysfunction, neurological disorders, tracheostomy, mechanical ventilation, central venous catheter (CVC) use, urinary catheter use, nasogastric tube use, implementation of dialysis, prior use of any antibiotic, and specific use of carbapenems, aminoglycosides, quinolones, fluoroquinolones, glycopeptides, or vancomycin. Table 4 shows the risk factors for CRKP infection for this comparison, as well as the heterogeneity in the metaanalysis. All 20 risk factors were dichotomous variables, and the following eight were statistically significant: admission to ICU, tracheostomy, mechanical ventilation, CVC use, urinary catheter use, prior antibiotic use, and specific use of carbapenems or aminoglycosides.

Publication bias
Egger's test showed no obvious asymmetry in the risk factors, suggesting low risk of publication bias (Tables 3  and 4).

Sensitivity analyses
The sensitivity analysis was performed by repeating the meta-analysis after omitting each study one by one and examining whether the results changed substantially. For most risk factors, no single study seemed to substantially alter the results. We noted two exceptions: in comparison 1, omitting the study by Mouloudi et al. from 2010 [30] made the factor "β-lactam + β-lactamase inhibitor" significant (OR 2.42, 95% CI 1.08 to 5.44); in comparison 2, removing the study by Mouloudi et al. in 2014 [37] made the factor "diabetes" significant (OR 1.39, 95% CI 1.01 to 1.90).

Discussion
CRKP is one of the most serious life-threating nosocomial pathogens worldwide, and CRKP infections are     [51]. The identification of risk factors of CRKP is the first step to discover high-risk patients and high-risk wards in  order to channel limited resources most effectively into prevention and treatment. Unfortunately, although many studies have investigated risk factors for CRKP infection, they have come to diverging, often conflicting, conclusions. For example, some studies have reported that exposure to carbapenems increased the risk of CRKP infection [17-22, 27, 29, 31, 33], but others did not find the same effect [24,30]. These discrepancies may reflect differences in sample size and overall lack of statistical power, which prompted us to perform a systematic review in order to assess the associations as reliably and comprehensively as possible.
We based our review on the idea that the choice of the control group for risk assessment can provide different results, as suggested in several previous studies [9][10][11][12]. We meta-analyzed 32 studies in nine countries involving several thousands of patients. Consistent with our initial idea, the profiles of risk factors differed between comparisons 1 and 2, with immediate implications for clinical practice. Comparison 1 assessed risk factors for carbapenemresistant infections, which are relevant for the situation when the patient is known to be infected with K. pneumoniae but tests of antibiotic susceptibility are pending. In this case, the clinician estimates the probability of resistance to carbapenem based on risk factors, adopting an empirical approach that prioritizes interventions to prevent transmission of carbapenem resistance at this early stage. In this type of comparison, our analysis identified the following risk factors: prior hospitalization (within the previous 6 months), longer length of stay, admission to the ICU, concomitant diseases (renal dysfunction, neurological disorders), certain invasive procedures (tracheostomy, mechanical ventilation, CVC, urinary catheter, nasogastric tube and dialysis), prior use of any antibiotic, and specific exposure to vancomycin or other five classes of antimicrobial agents (carbapenems, aminoglycosides, quinolones, fluoroquinolones, glycopeptides). These risk factors are more likely to be present in patients with more severe illness and greater susceptibility to infection, and who are therefore exposed to greater antibiotic selection pressure, which may ultimately increase the likelihood of infection with multidrug-resistant pathogens [20].
Comparison 2 is more relevant for the situation when hospitals need to identify patients at increased risk of suffering K. pneumoniae infection in general and CRKP in particular. The impact of risk factors on CRKP infection reflects an integrated effect of K. pneumoniae characteristics and carbapenem resistance. This may allow clinicians and hospital epidemiologists to take timely action to prevent CRKP transmission, even when no pathogen is detected in patient specimens, which may be due to their use of medications. In this type of comparison, our analysis identified the following risk factors: admission to ICU, certain invasive procedures (tracheostomy, mechanical ventilation, CVC, urinary catheter), prior use of any antibiotic, and exposure to carbapenems or aminoglycosides. Importantly, these risk factors were also statistically significant in comparison 1, which means that they are probably true risk factors for acquiring CRKP infection among hospitalized patients.
In contrast, dialysis and exposure to fluoroquinolones or glycopeptides were risk factors only for the first comparison. These factors may therefore increase primarily the risk of carbapenem resistance in K. pneumoniae. Indeed, fluoroquinolone exposure can generate resistance not only to fluoroquinolones but also to carbapenems, as fluoroquinolones lead to upregulation of the multidrug efflux pump MexEF-OprN and downregulation of the porin OprD, which is involved in carbapenem resistance [51,52]. In addition, a quinolone resistance gene that causes low-level fluoroquinolone resistance is located on K. pneumoniae plasmids carrying carbapenemase genes [52]. Long-term administration of the glycopeptide vancomycin may disrupt the balance of microflora in the body, promoting the propagation of Gram-negative bacteria and increasing the rate of mutation and spread of carbapenemases, which may augment the risk of CRKP [18]. These considerations imply that restricting the use of fluoroquinolones and glycopeptides, whenever possible, may decrease the transmission of carbapenem resistance.
Our sensitivity analysis confirmed that meta-analysis results were robust, with the possible exceptions of exposure to β-lactam + β-lactamase inhibitor (comparison 1) and diabetes (comparision 2). The status of these variables as risk factors changed depending on the inclusion of two small studies [30,37]. The heterogeneity surrounding these variables suggests the need for further studies to confirm their relationship with risk of CRKP infection.
Compared to a previous meta-analysis with a similar goal [8], the present work included 12 additional studies involving 2981 patients published after September 2016. In addition, we excluded studies comparing patients infected with CPKP with controls without CPKP infection, and our results for separate two comparisons contrast with a previous meta-analysis that aggregated both types of comparison. Consistent with our initial hypothesis, we identified several differences in the risk factors that were significant in each comparison, and we were able to derive a set of likely true risk factors of CRKP infection as those factors significant in both comparisons. The previous work identified the following significant risk factors: exposure to glycopeptides, parenteral nutrition, length of ICU stay and steroid therapy [8]. In our analysis, however, exposure to glycopeptides was significant only in comparison 1, while length of ICU stay and steroid therapy were not significant in comparison 1, and parenteral nutrition was not significant in either type of comparison, suggesting that these four factors may not be considered true risk factors. Furthermore, we found urinary catheter use to be a significant risk factor in both types of comparison, contrary to the previous metaanalysis.
Like the previously published meta-analysis on risk factors of CRKP infection [8], our exclusion criteria did not include that the source or base population of both case and control groups were identified with CRKP colonization based on rectal culture. With the exception of two studies [35,36], the studies included in our metaanalysis did not perform rectal screening for CRKP, and thus potential CRKP rectal colonization was not identified. In these cases, it was difficult to judge whether the risk factors associated with the process of CRKP colonization developing into infection or acquiring CRKP and having it cause infection. Moreover, the relative timing of CRKP colonization and onset of risk factors is often difficult to determine [36]. Further studies are needed in which risk factors associated with CRKP colonization developing into infection, which would then allow meta-analysis to identify the risk factors for CRKP infection among patients with CRKP colonization.
The findings of our meta-analysis should be interpreted with caution given that some potential risk factors were analyzed based on data from a small number of studies. Indeed, data for some factors showed significant heterogeneity across studies, especially in comparison 2, probably because control patients included those without any infection as well as those infected with nosocomial pathogens other than CRKP. Most studies in our review were retrospective and all were observational, increasing the risk of patient selection bias, outcome reporting bias and confounding. Nevertheless, all studies received NOS scores indicating high quality, and no obvious publication bias was observed for any of the factors. Factors affecting risk of CRKP infection should be further examined in large, well-controlled prospective studies.

Conclusions
This meta-analysis identified 18 factors that increase the risk of carbapenem resistance in K. pneumoniae infection and eight factors which were associated with both K. pneumoniae infections in general and CRKP in particular. The eight shared factors are probably 'true' risk factors for CRKP infection. These findings may help clinicians and hospital epidemiologists estimate the likelihood of CRKP infection in different situations, and thereby initiate timely, targeted treatment and prevention measures.